Intravenous immunoglobulin versus placebo

We planned to perform a meta‐analysis of the IVIg trials using data reported for our primary outcome, muscle strength, albeit at different time points. However, none of the IVIg studies reported data in a form that could be combined at 3, 6, or 12 months. Dalakas 1997 presented change in MRC sum scores for participants at three months in graphical form; we did not feel confident to accurately convert the graphical data to numerical values for meta‐analysis. Additionally, we expected the need to extrapolate three‐month data to six months to amplify any existing inaccuracies. In order to perform our analysis, we would have needed to rescale the original three‐month trial data to six months, assuming a linear rate of change in strength over time. Dalakas 2001 provided summary data for upper and lower limbs, but calculation of a percentage change from these summary scores and extrapolation to six months would again be subject to large assumptions in terms of the effect size. The Walter 2000 cross‐over trial similarly did not provide data in a form suitable for meta‐analysis.

Walter 2000 included 2 groups of 11 participants of different mean ages (67 years and 51 years); in theory, the cross‐over design of the trial should negate any baseline differences seen between the two groups. However, we were unable to perform subgroup analysis with respect to age because of insufficient data. Subsequently, we cannot rule out the influence of faster disease progression in the older subgroup in relation to sarcopenia and enhanced mitochondrial dysfunction, as proposed previously (Dalakas 2001; Santorelli 1996). Exploring the relationship between inflammation, mitochondrial dysfunction, and muscle atrophy and determining the rate of IBM progression could be essential for understanding any clinically relevant change with treatment (Rygiel 2014); additionally, we may need to consider the sexual dimorphic effect of ageing in relation to muscle protein synthesis in IBM, as postulated in the study of older adults (Smith 2012). While an ageing effect on our reviewed outcomes cannot be confirmed, the small, non‐significant improvements in muscle strength with both IVIg and placebo may indicate a general learning effect in strength measurement as a proxy measure for disease progression; Neuromuscular Symptom and Disability Functional Score changes showed statistically significant improvement with IVIg only. We planned to perform subgroup meta‐analyses for factors such as age and carry‐over effect, but insufficient data were available.

In terms of the intervention, Dalakas 2001, unlike the other two trials, provided the IVIg and non‐IVIg groups with equal doses of prednisone. The supplementation with prednisone was provided on the basis that the combination of steroid with IVIg might have a synergistic effect in improving muscle strength, based on studies in dermatomyositis and Guillain‐Barré syndrome (Dalakas 1993; Dutch GBS 1994). However, unlike for IBM, for dermatomyositis and Guillain‐Barré syndrome, IVIg alone is known to be effective. For the purposes of meta‐analysis, we would have to assume that prednisone did not contribute any effect for either the IVIg or non‐IVIg treatment group.

In terms of study design, the three trials used two different MRC scales, and they did not assess the same muscle groups or muscle group actions. Dalakas 1997 involved elective cross‐over in the second phase of the trial, such that we could consider only the first phase of intervention for meta‐analysis (Elbourne 2002). The inconsistency in methodology between trials is particularly important because IBM affects different muscles to a variable extent. As a result of the variable muscle involvement in IBM, IVIg might be beneficial for some muscles more than others at any given time point in the disease.

When summarising our findings, we assessed the quality of evidence for the effects of IVIg on muscle strength at six months as very low due to selective reporting and other issues of trial design (high risk of bias in one of the trials and an unclear overall risk of bias in the other two trials).

In conclusion, while marginal increments in muscle strength were identified in two of three trials of IVIg (Dalakas 1997; Walter 2000), we could not determine an overall effect of IVIg versus placebo due to inconsistencies in trial methodology and reporting (see summary of findings Table for the main comparison).

None of the IVIg trials included a statistical analysis of the incidence of significant adverse events with intervention to facilitate full evaluation of treatment effect. According to Meyler's Side Effects of Drugs (Chalker 2000; Dukes 2000), current Ig preparations cause about 3% to 4% of people to experience adverse reactions. Other adverse events associated with IVIg include mild influenza‐like illness, sweating, hypotension, chills, fever, nausea, and vasomotor reactions. More serious adverse effects include anaphylactic reactions; however, these are very rare and may occur in as few as 1 in 6000 people (Aronson 2006). Stroke and myocardial infarction have been reported after high‐dose IVIg, as a result of increased plasma viscosity. At high doses, neutropenia and disseminated intravascular coagulation have also been reported, and very rarely, acute renal failure (Aronson 2006). Walter 2000 reported some participants developing headache or raised body temperature, and two participants developed an allergic reaction.

Interferon beta‐1a versus placebo

Neither of the IFN beta‐1a studies showed a significantly greater benefit with IFN beta‐1a over placebo. We were able to perform a meta‐analysis of parallel‐group trials comparing different doses of IFN beta‐1a with placebo. The standardised trial procedures and reporting across both trials made pooling of data from Muscle Study Group 2001 and Muscle Study Group 2004 possible. The pooled analysis for normalised muscle strength change from baseline produced a MD in compound MMT in favour of placebo (moderate‐quality evidence), but with CIs including the possibility of an effect in either direction (Analysis 1.1; Figure 2; summary of findings Table 2). The forest plot for lean mass, as surrogate for muscle mass (Figure 3), also did not show an effect in favour of IFN beta‐1a (moderate‐quality evidence).

Despite some reported improvement in grip strength with IFN beta‐1a treatment, we could not assess the quality of this evidence because the available data were incomplete.

Neither of the IFN beta‐1a trials included a statistical analysis of the incidence of significant adverse events to facilitate full evaluation of the treatment effect. Among the participants who took 30 µg doses of IFN beta‐1a, 84% reported adverse events (Muscle Study Group 2001); this figure was approximately 81% in participants who took 60 µg doses of IFN beta‐1a (Muscle Study Group 2004). Flu‐like reactions following injection were the most commonly specified complaint with high‐dose treatment, while diarrhoea was the most commonly specified adverse event with low‐dose treatment (Muscle Study Group 2001; Muscle Study Group 2004). Rarer adverse events associated with IFN beta‐1a can include mood and personality changes, suicide attempts, hepatitis, and thyroid dysfunction (Aronson 2006). Such an adverse event profile and the large percentage of users experiencing more minor adverse events may give rise to problems with compliance over a longer period. However, despite the high frequency of adverse events seen in these two trials, none was assessed as significant enough to require dose reduction (Muscle Study Group 2001; Muscle Study Group 2004).

Methotrexate versus placebo

A single RCT provided moderate‐quality evidence of no significant effect of MTX on muscle strength at 12 months; the trial was insufficiently powered to exclude a possible benefit from MTX (see summary of findings Table 3). Approximately 38% of participants who were on MTX reported adverse events (Analysis 2.1), and there was a statistically significant dropout rate in the MTX treatment group (moderate‐quality evidence). A larger or longer trial of MTX could be problematic in terms of compliance (Badrising 2002). The most common adverse events caused by MTX are nausea, vomiting, alopecia, oral mucositis, or effects of myelosuppression (Aronson 2006). While there are pharmacological agents to reduce or alleviate common adverse events, polypharmacy may also affect compliance.

Combination therapy in the treatment of IBM

The benefit of combination therapy with MTX and ATG, Lindberg 2003, or MTX and AZA, Leff 1993, remains unclear in the treatment of IBM. Combined MTX and ATG appeared to show some benefit, but the evidence, based on a small, open‐label trial, is very low quality and should be interpreted with caution (see summary of findings Table 4) (Lindberg 2003). Aside from risk of bias issues, any positive effect of ATG combined with MTX versus MTX alone could either be due to the effects of ATG alone or to its effects when combined with MTX. Adverse effects known to be associated with ATG include "leukopenia and thrombocytopenia, fever, arthralgia, rash, urticaria, hepatotoxicity, hyperglycaemia, hypertension, and diarrhoea" (Aronson 2006). Serum sickness can also occur later. However, none of these potential adverse effects was reported in the trial; only a single case of pneumonia was reported.

In the only trial that looked at AZA and MTX in combination (Leff 1993), we could obtain no quantitative data (see summary of findings Table 5). Similar to Lindberg 2003, this trial was open label and had a small number of participants (very low‐quality evidence). As a combination therapy, there was a risk of exposure to the effects of both drugs in the combination. Adverse events on the regimen that included AZA were gastrointestinal symptoms, acute cholecystitis, and a case of pneumonitis. In the MTX alone group, the authors reported only a flareup of pre‐existing gout; other adverse events known to occur with AZA alone include fever, nausea, leukopenia, thrombocytopenia, and anaemia (Aronson 2006). Neither Lindberg 2003 nor Leff 1993 provided a statistical analysis of the incidence of significant adverse events with the intervention to facilitate a full evaluation of the treatment effect.

Oxandrolone versus placebo

In comparing oxandrolone with placebo, the trial authors highlighted that small numbers of participants and a relatively short trial duration restricted the interpretation of results (Rutkove 2002). The trial authors also considered that between‐group differences in baseline characteristics may have had some confounding effects. The trial authors reported a close‐to‐significant treatment effect for improving whole body strength with oxandrolone at 12 weeks. However, we noted that the methods of analysis (generalised estimating equation) were unusual for this trial design, and the power of the test for carry‐over effects was low. Taking into account these and the other study limitations, results should be interpreted with caution (very low‐quality evidence).

In terms of adverse events, anabolic steroids have some androgenic activity that can give rise to acne and other signs of virilisation and may affect lipoprotein profiles. As the associated androgenic activity is weak, these adverse events are not common. However, gynaecomastia has occurred with the long‐term use of anabolic steroids as a growth promoter in boys (Aronson 2006). Withdrawal of high doses of anabolic steroids can give rise to menopause‐like symptoms (Chalker 2000; Dukes 2000). Rutkove 2002 reported no significant adverse events, which was perhaps related to the low doses used over a relative short period of time but, as with all other analysed trials, the investigators included no definition of what constituted an adverse event in order to evaluate treatment effect. If anabolic steroids are associated with few or no significant adverse events in practice, possible therapeutic potential might justify further trials to explore their use in treating IBM. Conversely, more adverse events and issues with compliance may be anticipated in a longer trial of anabolic steroids.

Arimoclomol versus placebo

At the time of review, we did not attempt to evaluate the effects of arimoclomol for treating IBM because the relevant data were not available for systematic review. The trial was only available in abstract (Machado 2013), and it was powered to assess safety and tolerability, not treatment effect.

The lessons from these trials

This systematic review of treatment for IBM identified nine analysable RCTs and one RCT published in an abstract only. In terms of determining a treatment effect, only 2 of the included trials reported power calculations, and we evaluated all 10 trials as underpowered to detect a statistically significant effect. One trial that included power calculations was primarily a safety and tolerability study, rather than an efficacy study (Muscle Study Group 2001). The other trial indicated multiple reasons for the lower‐than‐expected power for their study, including rate of decline in the placebo group; variability in QMT strength measures; and a higher‐than‐expected participant dropout rate (Badrising 2002). The largest included RCT had only 22 participants in the treatment group (Walter 2000). The largest analysed treatment group included only 21 participants, using an intention‐to‐treat analysis (Badrising 2002), and 20 participants with a per protocol analysis of the primary outcome measure (Walter 2000). Rose 2001 calculated that each group of a placebo‐controlled trial needed 94 participantsto have 90% power to detect a 4% difference in mean change in muscle strength between two groups over a 6‐month period. This 4% larger difference over placebo was determined to be equivalent to arresting disease progression with the drug on trial. Based on trial experience, Muscle Study Group 2004 calculated that a 2‐arm, 6‐month intervention would require 208 participants per group to detect the arrest of disease progression using QMT with 90% power, a 2‐tailed t‐test, and significance level of 5%.

Multicentre recruitment, pooling of trial results, or both, might allow for the accumulation of sufficient data for a more robust answer as to the efficacy of treatment for IBM. However, we still expect differences in both methodologies and the type of intervention to affect treatment effect estimates and the validity of results. Across the studies included in this review, meta‐analysis was compromised by the fact that outcome measurement was not standardised to measure the cardinal effects of interest, namely muscle weakness, atrophy, and functional impairment. Additionally, some studies used normalised MDs for assessing change in muscle strength.

For muscle strength testing, it would be helpful to standardise appropriate test methodology (for example, manual or quantitative measurements (or both), specification and number of muscles or muscle actions tested, and the detection of a minimal clinically important difference) for use in research trials. The potential for harm resulting from treatment also needs to be carefully considered; none of the included studies defined what constituted an adverse effect, although drug treatments were associated with adverse effects of variable severity. Most of the studies also did not specify the methods used to monitor adverse effects. Analysis of treatment effect can risk bias towards a focus on favourable outcome measures in the absence of sufficient information on conduct and reporting of adverse events (Loke 2007).

None of the completed trials made reference to responsiveness or a minimal clinically important difference for any outcome measure in relation to disease progression. The validation of trial outcome measures, where reported, was also not performed specifically in an IBM population.

In terms of comparing future clinical trials in IBM, it will be important to minimise the differences in inclusion and exclusion criteria applied; this review found that trials used a range of criteria regarding comorbidity and concomitant treatment that could have a fundamental impact on study outcomes. Some have argued that insistence on pathological criteria for the inclusion of IBM in clinical trials may result in such trials attempting to treat participants who have more advanced, and therefore inherently less treatable, disease. Currently proposed diagnostic criteria reduce the emphasis on pathological criteria with the aim of allowing recruitment of participants earlier in the course of their disease. However, this strategy is subject to verification, and currently there are no trials that apply these newer criteria for IBM.

Costs

UK costs for treatment with the interventions used in the included studies would be:

• Interferon beta‐1a (Avonex): injection, 60 μg (12 million units)/mL, net price 0.5 mL (30 μg, 6 million‐unit) prefilled syringe = GBP 163.50 (BNF 2014).

• Methotrexate: tablets, 2.5 mg, net price 24‐tablet pack = GBP 2.22; 10 mg, net price 100‐tablet pack = GBP 37.06 (BNF 2014).

• Azathioprine: tablets, 50 mg, 56‐tablet pack = GBP 3.42 (BNF 2014).

• Intravenous immunoglobulin: based on 2 g/kg in 70 kg man = GBP 3906 to 4900 (DH 2011).

• Oxandrolone: a price from 2000 of USD 4 per 2.5 mg tablet (Beaston‐Blaakman 2007). In the UK, injection, nandrolone decanoate 50 mg/L, net price 1 mL ampoule = GBP 3.17 by deep intramuscular injection, 50 mg every 3 weeks (BNF 2014).

Intervention treatment costs per person over a six‐month period were estimated to be greatest for IVIg (approximately GBP 15,000) followed by IFN beta‐1a therapy, and with markedly lower costs for oxandrolone by intramuscular injection, AZA, and MTX treatment.

Potential biases in the review process

We chose to express the primary outcome as the percentage change in compound MMT over time because it is a widely used approach to muscle strength assessment in clinical research. However, percentage change calculation was rarely feasible using the available data, and two trials reported normalised data instead. Our percentage figures related to group‐level changes, but these are likely to be different from those calculated using absolute data from individuals. We expected the data to be subject to uncertainty in relation to clinical and statistical heterogeneity and significance. Using absolute changes would be preferable, but these were not retrievable across all trials in the review. Also, to calculate mean change in absolute MMT scores might have required that only those muscles tested by all studies and scaled to the same MRC grading were included in analysis.

Another potential bias in the review process was in the type of selected outcome measures. Our selected functional impairment outcome measures were restricted to specific single‐item tasks, while inflammatory and pathological biomarkers were not analysed at all. Nevertheless, our data extraction did suggest that included trials inconsistently assessed or reported those additional functional, inflammatory, and pathological outcomes.

Some of the review authors were investigators in included trials. A non‐conflicted review author performed independent data extraction if any review author had potential conflicts of interest, for example through involvement in an included study.

A further limitation of the review was that the review methods were unlikely to adequately detect serious, rare adverse events. We therefore discussed adverse events described in other sources in the Discussion.

Agreements and disagreements with other studies or reviews

The results of this systematic review confirmed previous observational study and review findings that there are no established, evidence‐based treatments for IBM as yet (Benveniste 2011; Breithaupt 2013; Machado 2013b). The small sample sizes and short duration of clinical trials are recognised as major limitations in the evaluation of treatment efficacy in this muscle condition (Breithaupt 2013; Fergusson 2005). In terms of assessment, MMT scores of muscle strength are widely used in research and clinical practice, but further development of sensitive outcome measures is advocated (Breithaupt 2013; Machado 2013). Studies included in this review used different methodological and analytical approaches to assess muscle strength and applied a range of secondary outcome measures. Such variations in methods and outcome measures hindered the pooling of trial data to contribute to the evidence base. As part of a systematic review in neurological conditions, Fergusson 2005 analysed the same three IVIg trials for IBM as in our review; like us, they were unable to come to a conclusion about the efficacy of IVIg from the available data. Fergusson 2005 similarly elected not to perform pooled analysis due to between‐trial differences in methodology. Interestingly, the benefit of IVIg in chronic inflammatory demyelinating polyneuropathy (CIDP) was identified using a disability scale, as was the benefit Walter 2000 showed for IVIg in IBM, using the Neuromuscular Symptom and Disability Functional Score. Such evidence perhaps argues for the use of disability scales in future IBM trials. However, Fergusson 2005 also emphasised that evidence of benefit still does not necessarily support IVIg as a first‐line treatment owing to other factors, including adverse events and cost.

Evidence of mitochondrial abnormalities in IBM that are in excess of those seen with normal ageing may suggest another range of therapeutic options, but there are currently no known effective treatments for primary mitochondrial disease (Pfeffer 2012). Similarly, currently there are no drugs with the therapeutic potential to arrest or slow down the degenerative pathology seen in IBM (Breithaupt 2013). A common issue highlighted across Cochrane intervention reviews in muscle disease is the quality of study design and paucity of RCTs, emphasising ongoing problems with sample size and risk of bias (Hill 2004; Pfeffer 2012; Voet 2013). The validity of the evidence base also appears to be limited by a lack of standardisation in the collection, evaluation, and reporting of data. This systematic review identified specific quality issues in clinical trials of people with IBM, which should encourage investigators to validate outcome measures and ensure standardisation of trial procedures.

Only one of the studies included in this review measured swallowing function, which we did not include as a predefined outcome measure. In the future a Cochrane review of treatment for swallowing difficulties in chronic muscle disease will include IBM (Hill 2004). In terms of ongoing drug trials for IBM, we identified an RCT of bimagrumab (BYM338), which has been developed to target molecular pathways involved in muscle growth, and which we may include in a future review. An RCT of simvastatin is also pending completion.